A Position- and Similarity-Aware Named Entity Recognition Model for Power Equipment Maintenance Work Orders
Abstract
:1. Introduction
- We construct the power equipment maintenance work orders (PE-MWO) dataset. The dataset comprises seven entity categories, encompassing a total of 7415 sentences and 238,869 characters.
- We develop a position- and similarity-aware attention module that refines the position embedding and attention score computation processes of the original Transformer model. Additionally, this module incorporates vector angle calculations to enhance the model’s capacity to detect token similarity information. This modification enables more nuanced contextual understanding in natural language processing tasks.
- Extensive experimental results provide compelling evidence for the efficacy of our proposed model. It demonstrates high recognition accuracy across both the PE-MWO dataset and five public datasets. This approach significantly advances the resolution of power industry-specific NER challenges, offering a robust reference for analogous domain-specific research.
2. Related Work
2.1. AI Applications in Power and Energy Management
2.2. Named Entity Recognition Methods
2.3. Core Components of Transformer
2.3.1. Self-Attention
2.3.2. Linear Attention
2.3.3. Position Embedding
3. Proposed Method
3.1. Character Embedding
3.2. Position- and Similarity-Aware Attention Module
3.2.1. Position-Aware
3.2.2. Similarity-Aware
3.3. CRF
4. Experiments
4.1. Datasets and Experimental Settings
4.1.1. PE-MWO Dataset
- The text length of different work orders varies significantly. In our collected data, the shortest text is 9 characters, while the longest text is 362 characters.
- Since maintenance work orders are written by personnel with different writing habits, the writing format of different work orders varies, and there are differences in the descriptions of the same issue.
- The maintenance work order texts involve a wide variety of entities, specialized vocabulary, abbreviated vocabulary, semantic complexity, and unclear segmentation boundaries.
4.1.2. Public Datasets
- MSRA is a dataset released by Microsoft Research Asia, widely used in the field of natural language processing. This dataset contains a large amount of Chinese text from news reports annotated with named entities, including entity types such as person names, location names, and organization names.
- Resume is specifically designed for NER tasks. It contains a large number of resume texts from various industries and fields, covering a wide range of professions and positions. The dataset includes a total of eight entity categories: country, educational background, location, name, organization, profession, ethnicity, and title.
- OntoNotes 4.0 is a multilingual, multi-domain dataset widely used for natural language processing tasks, covering various domains such as news, conversations, and the web. In this experiment, we selected the Chinese NER portion of the dataset, which includes four entity categories: person, organization, location, and geopolitical entity.
- China People’s Daily corpus, derived from China’s authoritative media outlets, is characterized by its standardized linguistic patterns and formal rhetorical conventions. This comprehensive dataset encompasses multiple domains, including political affairs, economic developments, cultural matters, and societal issues, constituting an extensive repository of domain-specific terminology and nomenclature that exemplifies the practical challenges in named entity recognition research.
- CoNLL-2003 is a benchmark for named entity recognition (NER) and contains English news articles annotated with four entity types: PER (person), LOC (location), ORG (organization), and MISC (miscellaneous). It includes labeled training, development, and test sets, all of which are formatted with word-level BIO tagging. English data derives from the Reuters corpus, supporting model training and evaluation in NLP tasks.
4.1.3. Evaluation Metrics
4.1.4. Hyperparameter Settings and Computational Resources
4.2. Ablation Study
4.3. Visual Analysis
4.3.1. Confusion Matrix Heatmaps
- Long-distance dependency problems: Although the model is able to capture contextual information, there may still be the issue that it cannot fully capture long-distance dependencies for some entity categories, especially those spanning longer sentences. This is particularly important for entity classes such as DamagePart and MaintenanceStatus, whose descriptions are often long and may be semantically intertwined with other entities, making it more difficult for the model to capture their relationships.
- The problem of polysemy and synonymy: In the context of power equipment maintenance, certain words are polysemous and may share vocabulary with multiple entity categories. For example, Substation may have some semantic overlap with Line or EquipmentName, leading to unstable model categorization between these categories. Even though the model mitigates this problem to some extent through location and similarity-aware mechanisms, the model may still produce misclassification in some cases due to the lack of deeper semantic understanding.
- High-dimensional input features: Although BERT-wwm-ext is able to provide effective feature representation for Chinese text, when the dimensionality of the input features is too high (especially in long text or complex entity descriptions), the model may capture too many irrelevant features, which in turn affects the prediction accuracy.
4.3.2. Comparative Analysis of Prediction Instances
4.4. Overall Performance Comparison
5. Conclusions and Future Perspectives
- Sensor Data Integration: Our structured maintenance information could be integrated with sensor measurements to provide more comprehensive equipment health monitoring. This integration would combine human expert knowledge from maintenance records with quantitative sensor data.
- Sensor-Based Validation: Future research could explore using sensor measurements to validate and enrich the extracted maintenance information, helping to establish more reliable equipment status assessment systems.
- Security Risk Assessment: The extracted structured information could be utilized to develop more sophisticated power equipment security risk assessment models, helping to identify potential vulnerabilities and prevent equipment failures.
- Smart Grid Applications: Future work could focus on incorporating the extracted maintenance information into smart grid management systems, supporting more intelligent decision-making in power equipment operation and maintenance scheduling.
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
- Dileep, G. A survey on smart grid technologies and applications. Renew. Energy 2020, 146, 2589–2625. [Google Scholar] [CrossRef]
- Mbungu, N.T.; Ismail, A.A.; AlShabi, M.; Bansal, R.C.; Elnady, A.; Hamid, A.K. Control and estimation techniques applied to smart microgrids: A review. Renew. Sustain. Energy Rev. 2023, 179, 113251. [Google Scholar] [CrossRef]
- Li, J.; Sun, A.; Han, J.; Li, C. A survey on deep learning for named entity recognition. IEEE Trans. Knowl. Data Eng. 2020, 34, 50–70. [Google Scholar] [CrossRef]
- Ji, S.; Pan, S.; Cambria, E.; Marttinen, P.; Philip, S.Y. A survey on knowledge graphs: Representation, acquisition, and applications. IEEE Trans. Neural Netw. Learn. Syst. 2021, 33, 494–514. [Google Scholar] [CrossRef] [PubMed]
- Jin, Q.; Yuan, Z.; Xiong, G.; Yu, Q.; Ying, H.; Tan, C.; Chen, M.; Huang, S.; Liu, X.; Yu, S. Biomedical question answering: A survey of approaches and challenges. ACM Comput. Surv. 2022, 55, 1–36. [Google Scholar] [CrossRef]
- Collobert, R.; Weston, J.; Bottou, L.; Karlen, M.; Kavukcuoglu, K.; Kuksa, P. Natural language processing (almost) from scratch. J. Mach. Learn. Res. 2011, 12, 2493–2537. [Google Scholar]
- Huang, Z.; Xu, W.; Yu, K. Bidirectional LSTM-CRF models for sequence tagging. arXiv 2015, arXiv:1508.01991. [Google Scholar]
- Li, R.; Mo, T.; Yang, J.; Li, D.; Jiang, S.; Wang, D. Bridge inspection named entity recognition via BERT and lexicon augmented machine reading comprehension neural model. Adv. Eng. Inform. 2021, 50, 101416. [Google Scholar] [CrossRef]
- Liu, C.; Yang, S. A text mining-based approach for understanding Chinese railway incidents caused by electromagnetic interference. Eng. Appl. Artif. Intell. 2023, 117, 105598. [Google Scholar] [CrossRef]
- Jumper, J.; Evans, R.; Pritzel, A.; Green, T.; Figurnov, M.; Ronneberger, O.; Tunyasuvunakool, K.; Bates, R.; Žídek, A.; Potapenko, A.; et al. Highly accurate protein structure prediction with AlphaFold. Nature 2021, 596, 583–589. [Google Scholar] [CrossRef]
- Nepal, U.; Eslamiat, H. Comparing YOLOv3, YOLOv4 and YOLOv5 for autonomous landing spot detection in faulty UAVs. Sensors 2022, 22, 464. [Google Scholar] [CrossRef] [PubMed]
- Srinivasu, P.N.; SivaSai, J.G.; Ijaz, M.F.; Bhoi, A.K.; Kim, W.; Kang, J.J. Classification of skin disease using deep learning neural networks with MobileNet V2 and LSTM. Sensors 2021, 21, 2852. [Google Scholar] [CrossRef] [PubMed]
- Gomes, B.; Ashley, E.A. Artificial intelligence in molecular medicine. N. Engl. J. Med. 2023, 388, 2456–2465. [Google Scholar]
- Huang, W.; Wei, Z.; Tan, B.; Yin, S.; Zhang, W. Inverse engineering of electromagnetically induced transparency in terahertz metamaterial via deep learning. J. Phys. D Appl. Phys. 2021, 54, 135102. [Google Scholar]
- Yin, S.; Zhong, H.; Huang, W.; Zhang, W.; Han, J. Deep Learning Enabled Design of Terahertz High-Q Metamaterials. arXiv 2023, arXiv:2312.13986. [Google Scholar]
- Wang, C.-f.; Liu, K.-x.; Peng, J.; Li, X.; Liu, X.-f.; Zhang, J.-w.; Niu, Z.-b. High-precision energy consumption forecasting for large office building using a signal decomposition-based deep learning approach. Energy 2025, 314, 133964. [Google Scholar]
- Qiu, H.; Zhang, Q.; Wang, R.; Liu, X.; Cheng, X.; Wu, L. A novel two-enhancive aspect module in convolutional neural networks for multivariate time series classification. Expert Syst. Appl. 2025, 266, 125755. [Google Scholar]
- Wen, H.; Liu, X.; Lei, B.; Yang, M.; Cheng, X.; Chen, Z. A privacy-preserving heterogeneous federated learning framework with class imbalance learning for electricity theft detection. Appl. Energy 2025, 378, 124789. [Google Scholar]
- Peng, J.; Kimmig, A.; Wang, D.; Niu, Z.; Liu, X.; Tao, X.; Ovtcharova, J. Energy consumption forecasting based on spatio-temporal behavioral analysis for demand-side management. Appl. Energy 2024, 374, 124027. [Google Scholar]
- Razzak, A.; Islam, M.T.; Roy, P.; Razzaque, M.A.; Hassan, M.R.; Hassan, M.M. Leveraging deep q-learning to maximize consumer quality of experience in smart grid. Energy 2024, 290, 130165. [Google Scholar]
- Zhang, L.; Xia, P.; Ma, X.; Yang, C.; Ding, X. Enhanced Chinese named entity recognition with multi-granularity BERT adapter and efficient global pointer. Complex Intell. Syst. 2024, 10, 4473–4491. [Google Scholar]
- Yan, H.; Deng, B.; Li, X.; Qiu, X. TENER: Adapting transformer encoder for named entity recognition. arXiv 2019, arXiv:1911.04474. [Google Scholar]
- Cheng, J.; Liu, J.; Xu, X.; Xia, D.; Liu, L.; Sheng, V.S. A review of Chinese named entity recognition. KSII Trans. Internet Inf. Syst. 2021, 15, 2012–2030. [Google Scholar]
- Liu, Y.; Huang, S.; Li, R.; Yan, N.; Du, Z. USAF: Multimodal Chinese named entity recognition using synthesized acoustic features. Inf. Process. Manag. 2023, 60, 103290. [Google Scholar]
- Zhang, Q.; Xue, C.; Su, X.; Zhou, P.; Wang, X.; Zhang, J. Named entity recognition for Chinese construction documents based on conditional random field. Front. Eng. Manag. 2023, 10, 237–249. [Google Scholar]
- Vaswani, A.; Shazeer, N.; Parmar, N.; Uszkoreit, J.; Jones, L.; Gomez, A.N.; Kaiser, Ł.; Polosukhin, I. Attention is all you need. Adv. Neural Inf. Process. Syst. 2017, 30, 1–11. [Google Scholar]
- Lauriola, I.; Lavelli, A.; Aiolli, F. An introduction to deep learning in natural language processing: Models, techniques, and tools. Neurocomputing 2022, 470, 443–456. [Google Scholar] [CrossRef]
- Acheampong, F.A.; Nunoo-Mensah, H.; Chen, W. Transformer models for text-based emotion detection: A review of BERT-based approaches. Artif. Intell. Rev. 2021, 54, 5789–5829. [Google Scholar]
- Bello, A.; Ng, S.C.; Leung, M.F. A BERT framework to sentiment analysis of tweets. Sensors 2023, 23, 506. [Google Scholar] [CrossRef]
- Xu, P.; Zhu, X.; Clifton, D.A. Multimodal learning with transformers: A survey. IEEE Trans. Pattern Anal. Mach. Intell. 2023, 45, 12113–12132. [Google Scholar] [CrossRef]
- Zhang, H.; Song, H.; Li, S.; Zhou, M.; Song, D. A survey of controllable text generation using transformer-based pre-trained language models. ACM Comput. Surv. 2023, 56, 1–37. [Google Scholar] [CrossRef]
- Brauwers, G.; Frasincar, F. A survey on aspect-based sentiment classification. ACM Comput. Surv. 2022, 55, 1–37. [Google Scholar] [CrossRef]
- Mehrish, A.; Majumder, N.; Bharadwaj, R.; Mihalcea, R.; Poria, S. A review of deep learning techniques for speech processing. Inf. Fusion 2023, 99, 101869. [Google Scholar] [CrossRef]
- Mei, S.; Song, C.; Ma, M.; Xu, F. Hyperspectral image classification using group-aware hierarchical transformer. IEEE Trans. Geosci. Remote Sens. 2022, 60, 1–14. [Google Scholar] [CrossRef]
- Wang, G.; Chen, Y.; An, P.; Hong, H.; Hu, J.; Huang, T. UAV-YOLOv8: A small-object-detection model based on improved YOLOv8 for UAV aerial photography scenarios. Sensors 2023, 23, 7190. [Google Scholar] [CrossRef]
- Shamshad, F.; Khan, S.; Zamir, S.W.; Khan, M.H.; Hayat, M.; Khan, F.S.; Fu, H. Transformers in medical imaging: A survey. Med Image Anal. 2023, 88, 102802. [Google Scholar] [CrossRef] [PubMed]
- Katharopoulos, A.; Vyas, A.; Pappas, N.; Fleuret, F. Transformers are rnns: Fast autoregressive transformers with linear attention. In Proceedings of the International Conference on Machine Learning, PMLR, Virtual, 13–18 July 2020; pp. 5156–5165. [Google Scholar]
- Tay, Y.; Dehghani, M.; Abnar, S.; Shen, Y.; Bahri, D.; Pham, P.; Rao, J.; Yang, L.; Ruder, S.; Metzler, D. Long range arena: A benchmark for efficient transformers. arXiv 2020, arXiv:2011.04006. [Google Scholar]
- Ma, X.; Kong, X.; Wang, S.; Zhou, C.; May, J.; Ma, H.; Zettlemoyer, L. Luna: Linear unified nested attention. Adv. Neural Inf. Process. Syst. 2021, 34, 2441–2453. [Google Scholar]
- You, H.; Xiong, Y.; Dai, X.; Wu, B.; Zhang, P.; Fan, H.; Vajda, P.; Lin, Y.C. Castling-vit: Compressing self-attention via switching towards linear-angular attention at vision transformer inference. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Vancouver, BC, Canada, 18–22 June 2023; pp. 14431–14442. [Google Scholar]
- de Santana Correia, A.; Colombini, E.L. Attention, please! A survey of neural attention models in deep learning. Artif. Intell. Rev. 2022, 55, 6037–6124. [Google Scholar] [CrossRef]
- Yan, X.; Fang, Z.; Jin, Y. An adaptive n-gram transformer for multi-scale scene text recognition. Knowl.-Based Syst. 2023, 280, 110964. [Google Scholar] [CrossRef]
- Ding, L.; Ouyang, C.; Liu, Y.; Tao, Z.; Wan, Y.; Gao, Z. Few-shot Named Entity Recognition via encoder and class intervention. AI Open 2024, 5, 39–45. [Google Scholar]
- Kim, H.; Kim, H. Recursive label attention network for nested named entity recognition. Expert Syst. Appl. 2024, 249, 123657. [Google Scholar] [CrossRef]
- Church, K.W. Word2Vec. Nat. Lang. Eng. 2017, 23, 155–162. [Google Scholar]
- Wang, W.; Gang, J. Application of convolutional neural network in natural language processing. In Proceedings of the 2018 International Conference on Information Systems and Computer Aided Education (ICISCAE), Changchun, China, 6–8 July 2018; pp. 64–70. [Google Scholar]
- Rogers, A.; Kovaleva, O.; Rumshisky, A. A primer in BERTology: What we know about how BERT works. Trans. Assoc. Comput. Linguist. 2021, 8, 842–866. [Google Scholar]
- Devlin, J.; Chang, M.W.; Lee, K.; Toutanova, K. Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv 2018, arXiv:1810.04805. [Google Scholar]
- Cui, Y.; Che, W.; Liu, T.; Qin, B.; Yang, Z. Pre-training with whole word masking for chinese bert. IEEE/ACM Trans. Audio Speech Lang. Process. 2021, 29, 3504–3514. [Google Scholar]
- Dai, Z.; Yang, Z.; Yang, Y.; Carbonell, J.; Le, Q.V.; Salakhutdinov, R. Transformer-xl: Attentive language models beyond a fixed-length context. arXiv 2019, arXiv:1901.02860. [Google Scholar]
- Liu, L.; Cai, L.; Zhang, C.; Zhao, X.; Gao, J.; Wang, W.; Lv, Y.; Fan, W.; Wang, Y.; He, M.; et al. Linrec: Linear attention mechanism for long-term sequential recommender systems. In Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval, Taipei, Taiwan, 23–27 July 2023; pp. 289–299. [Google Scholar]
- Han, D.; Pan, X.; Han, Y.; Song, S.; Huang, G. Flatten transformer: Vision transformer using focused linear attention. In Proceedings of the IEEE/CVF International Conference on Computer Vision, Paris, France, 2–10 October 2023; pp. 5961–5971. [Google Scholar]
- Daubechies, I.; DeVore, R.; Foucart, S.; Hanin, B.; Petrova, G. Nonlinear approximation and (deep) ReLU networks. Constr. Approx. 2022, 55, 127–172. [Google Scholar]
- You, H.; Shi, H.; Guo, Y.; Lin, Y. ShiftAddViT: Mixture of multiplication primitives towards efficient vision transformer. Adv. Neural Inf. Process. Syst. 2024, 36, 33319–33337. [Google Scholar]
- Lafferty, J.; McCallum, A.; Pereira, F. Conditional random fields: Probabilistic models for segmenting and labeling sequence data. In Proceedings of the Eighteenth International Conference on Machine Learning, Williamstown, MA, USA, 28 June–1 July 2001; Volume 1, pp. 282–289. [Google Scholar]
- Levow, G.A. The third international Chinese language processing bakeoff: Word segmentation and named entity recognition. In Proceedings of the Fifth SIGHAN Workshop on Chinese Language Processing, Sydney, Australia, 22–23 July 2006; pp. 108–117. [Google Scholar]
- Zhang, Y.; Yang, J. Chinese NER using lattice LSTM. arXiv 2018, arXiv:1805.02023. [Google Scholar]
- Che, W.; Wang, M.; Manning, C.D.; Liu, T. Named entity recognition with bilingual constraints. In Proceedings of the 2013 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Atlanta, GA, USA, 9–15 June 2013; pp. 52–62. [Google Scholar]
- Sang, E.F.; De Meulder, F. Introduction to the CoNLL-2003 shared task: Language-independent named entity recognition. arXiv 2003, arXiv:cs/0306050. [Google Scholar]
- Zhang, N.; Li, F.; Xu, G.; Zhang, W.; Yu, H. Chinese NER using dynamic meta-embeddings. IEEE Access 2019, 7, 64450–64459. [Google Scholar] [CrossRef]
- Guo, X.; Zhou, H.; Su, J.; Hao, X.; Tang, Z.; Diao, L.; Li, L. Chinese agricultural diseases and pests named entity recognition with multi-scale local context features and self-attention mechanism. Comput. Electron. Agric. 2020, 179, 105830. [Google Scholar]
- Gong, C.; Li, Z.; Xia, Q.; Chen, W.; Zhang, M. Hierarchical LSTM with char-subword-word tree-structure representation for Chinese named entity recognition. Sci. China Inf. Sci. 2020, 63, 1–15. [Google Scholar]
- Li, D.; Yan, L.; Yang, J.; Ma, Z. Dependency syntax guided bert-bilstm-gam-crf for chinese ner. Expert Syst. Appl. 2022, 196, 116682. [Google Scholar] [CrossRef]
- Chen, J.; Xi, X.; Sheng, V.S.; Cui, Z. Randomly wired graph neural network for chinese ner. Expert Syst. Appl. 2023, 227, 120245. [Google Scholar] [CrossRef]
- Ke, X.; Wu, X.; Ou, Z.; Li, B. Chinese Named Entity Recognition method based on multi-feature fusion and biaffine. Complex Intell. Syst. 2024, 10, 6305–6318. [Google Scholar] [CrossRef]
- Han, Q.; Ma, J. A Chinese named entity recognition model incorporating recurrent cell and information state recursion. Sci. Rep. 2024, 14, 5564. [Google Scholar] [CrossRef]
- Wang, Y.; Lu, L.; Wu, Y.; Chen, Y. Polymorphic graph attention network for Chinese NER. Expert Syst. Appl. 2022, 203, 117467. [Google Scholar]
- Wang, Q.; Wen, Z.; Ding, K.; Zhao, Q.; Yang, M.; Yu, X.; Xu, R. Improving sequence labeling with labeled clue sentences. Knowl.-Based Syst. 2022, 257, 109828. [Google Scholar]
- Tian, X.; Bu, X.; He, L. Multi-task learning with helpful word selection for lexicon-enhanced Chinese NER. Appl. Intell. 2023, 53, 19028–19043. [Google Scholar]
- Mai, C.; Liu, J.; Qiu, M.; Luo, K.; Peng, Z.; Yuan, C.; Huang, Y. Pronounce differently, mean differently: A multi-tagging-scheme learning method for Chinese NER integrated with lexicon and phonetic features. Inf. Process. Manag. 2022, 59, 103041. [Google Scholar]
- Zhang, L.; Nie, X.; Zhang, M.; Gu, M.; Geissen, V.; Ritsema, C.J.; Niu, D.; Zhang, H. Lexicon and attention-based named entity recognition for kiwifruit diseases and pests: A Deep learning approach. Front. Plant Sci. 2022, 13, 1053449. [Google Scholar] [CrossRef] [PubMed]
- Runmei, Z.; Lulu, L.; Lei, Y.; Jingjing, L.; Weiyi, X.; Weiwei, C.; Zhong, C. Chinese named entity recognition method combining ALBERT and a local adversarial training and adding attention mechanism. Int. J. Semant. Web Inf. Syst. 2022, 18, 1–20. [Google Scholar]
- Liu, J.; Sun, M.; Zhang, W.; Xie, G.; Jing, Y.; Li, X.; Shi, Z. DAE-NER: Dual-channel attention enhancement for Chinese named entity recognition. Comput. Speech Lang. 2023, 85, 101581. [Google Scholar]
- Yi, Y.; Bian, Y. Named entity recognition with gating mechanism and parallel BiLSTM. J. Web Eng. 2021, 20, 1219–1238. [Google Scholar]
- Chen, Y.; Wu, C.; Qi, T.; Yuan, Z.; Zhang, Y.; Yang, S.; Guan, J.; Sun, D.; Huang, Y. Semi-supervised named entity recognition in multi-level contexts. Neurocomputing 2023, 520, 194–204. [Google Scholar]
- Fei, Y.; Xu, X. GFMRC: A machine reading comprehension model for named entity recognition. Pattern Recognit. Lett. 2023, 172, 97–105. [Google Scholar]
- Chang, J.; Han, X. Multi-level context features extraction for named entity recognition. Comput. Speech Lang. 2023, 77, 101412. [Google Scholar]
- Yu, Y.; Wang, Z.; Wei, W.; Zhang, R.; Mao, X.L.; Feng, S.; Wang, F.; He, Z.; Jiang, S. Exploiting global contextual information for document-level named entity recognition. Knowl.-Based Syst. 2024, 284, 111266. [Google Scholar] [CrossRef]
Entity Category | Entity Meaning | Entity Quantity |
---|---|---|
EquipmentName | The names of the equipment in the maintenance work orders | 4184 |
VoltageLevel | The voltage levels of the equipment | 2802 |
Line | The lines to which the equipment belongs | 1496 |
Substation | The substations to which the equipment belongs | 1474 |
DamagePart | The damaged parts of the equipment | 2705 |
MaintenanceStatus | The specific maintenance details of the equipment | 4514 |
Time | The maintenance time of the equipment | 228 |
Dataset | Type | Train (k) | Dev (k) | Test (k) | Entity Type |
---|---|---|---|---|---|
MSRA | Sentences | 46.4 | 4.4 | 4.4 | 3 |
Chars | 2169.9 | 172.6 | 172.6 | ||
Entities | 74.8 | 6.2 | 6.2 | ||
Resume | Sentences | 3.8 | 0.46 | 0.48 | 8 |
Chars | 124.1 | 13.9 | 15.1 | ||
Entities | 13.4 | 1.5 | 1.6 | ||
OntoNote 4.0 | Sentences | 15.7 | 4.3 | 4.3 | 4 |
Chars | 491.9 | 200.5 | 208.1 | ||
Entities | 13.3 | 6.9 | 7.7 | ||
People’s Daily | Sentences | 20.86 | 2.32 | 4.64 | 3 |
Chars | 979.2 | 109.9 | 219.2 | ||
Entities | 33.9 | 3.8 | 7.7 | ||
CoNLL-2003 | Sentences | 15.0 | 3.4 | 3.6 | 4 |
Chars | 203.6 | 51.4 | 46.4 | ||
Entities | 23.5 | 5.9 | 5.6 |
Hyperparameters | MSRA | Other Datasets |
---|---|---|
Number of layers | 2 | 2 |
Number of head | 6 | 6 |
Optimizer | Adam | Adam |
Learning rate | 0.0007 | 0.0009 |
Batch size | 16 | 64 |
Dropout | 0.4 | 0.4 |
Epochs | 50 | 50 |
Configuration Item | Specifications/Model |
---|---|
GPU | NVIDIA GeForce RTX 4090 |
CPU | Intel Core i9-14900KF |
Operating System | Windows 10 Pro |
Deep Learning Framework | PyTorch 2.1.2 + CUDA 12.1 |
Python Environment | Python 3.10 |
Model | P (%) | R (%) | F1 (%) |
---|---|---|---|
Baseline 1 | 76.9 | 75.9 | 76.4 |
+Position-Aware 2 | 82.9 | 78.6 | 80.7 |
+Similarity-Aware 3 | 76.8 | 82.6 | 79.6 |
+Position- and Similarity-Aware 4 | 84.9 | 89.0 | 86.9 |
+All 5 | 91.5 | 92.1 | 91.8 |
Model | P (%) | R (%) | F1 (%) |
---|---|---|---|
Zhang et al., 2019 [60] | 90.59 | 91.15 | 90.87 |
Guo et al., 2020 [61] | 89.76 | 91.82 | 90.55 |
Gong et al., 2020 [62] | 94.83 | 93.61 | 94.22 |
Li et al., 2022 [63] | 95.00 | 94.96 | 94.97 |
Chen et al., 2023 [64] | 96.84 | 93.78 | 95.29 |
Ke et al., 2024 [65] | 96.28 | 96.23 | 96.26 |
Han et al., 2024 [66] | 95.83 | 96.84 | 96.34 |
Ours | 97.08 | 95.71 | 96.39 |
Model | P (%) | R (%) | F1 (%) |
---|---|---|---|
Gong et al., 2020 [62] | 77.77 | 76.32 | 77.04 |
Wang et al., 2022 [67] | - | - | 81.87 |
Wang et al., 2022 [68] | - | - | 82.43 |
Chen et al., 2023 [64] | 77.64 | 76.26 | 76.94 |
Tian et al., 2023 [69] | 82.61 | 84.29 | 83.44 |
Han et al., 2024 [66] | 82.16 | 83.51 | 82.83 |
Zhang et al., 2024 [21] | - | - | 83.25 |
Ours | 82.59 | 86.27 | 84.39 |
Model | P (%) | R (%) | F1 (%) |
---|---|---|---|
Guo et al., 2020 [61] | 94.9 | 94.56 | 94.62 |
Wang et al., 2022 [68] | - | - | 96.21 |
Wang et al., 2022 [67] | - | - | 96.53 |
Mai et al., 2022 [70] | 96.91 | 96.26 | 96.58 |
Chen et al., 2023 [64] | 96.14 | 96.52 | 96.33 |
Tian et al., 2023 [69] | 96.69 | 96.81 | 96.75 |
Zhang et al., 2024 [21] | - | - | 96.23 |
Ours | 96.4 | 97.17 | 96.78 |
Model | P (%) | R (%) | F1 (%) |
---|---|---|---|
Zhang et al., 2022 [71] | 93.23 | 92.42 | 92.82 |
Zhang et al., 2022 [72] | 94.71 | 92.42 | 93.64 |
Liu et al., 2024 [73] | 92.17 | 90.63 | 91.4 |
Ke et al., 2024 [65] | 95.93 | 96.45 | 96.19 |
Ours | 97.53 | 96.38 | 96.95 |
Model | P (%) | R (%) | F1 (%) |
---|---|---|---|
Yi et al., 2021 [74] | 91.18 | 91.36 | 91.27 |
Chen et al., 2023 [75] | - | - | 93.09 |
Fei et al., 2023 [76] | 92.96 | 93.85 | 93.4 |
Chang et al., 2023 [77] | - | - | 93.46 |
Yu et al., 2024 [78] | - | - | 93.42 |
Ours | 94.05 | 93.76 | 93.91 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Wei, Z.; Qu, S.; Zhao, L.; Shi, Q.; Zhang, C. A Position- and Similarity-Aware Named Entity Recognition Model for Power Equipment Maintenance Work Orders. Sensors 2025, 25, 2062. https://doi.org/10.3390/s25072062
Wei Z, Qu S, Zhao L, Shi Q, Zhang C. A Position- and Similarity-Aware Named Entity Recognition Model for Power Equipment Maintenance Work Orders. Sensors. 2025; 25(7):2062. https://doi.org/10.3390/s25072062
Chicago/Turabian StyleWei, Ziming, Shaocheng Qu, Li Zhao, Qianqian Shi, and Chen Zhang. 2025. "A Position- and Similarity-Aware Named Entity Recognition Model for Power Equipment Maintenance Work Orders" Sensors 25, no. 7: 2062. https://doi.org/10.3390/s25072062
APA StyleWei, Z., Qu, S., Zhao, L., Shi, Q., & Zhang, C. (2025). A Position- and Similarity-Aware Named Entity Recognition Model for Power Equipment Maintenance Work Orders. Sensors, 25(7), 2062. https://doi.org/10.3390/s25072062